This turned out to be that annoying bug that haunted my DCOM work for so
long...
Anyway, now that we've finally gotten it to work, I plan to submit my
DCOM work to Wine piecewise over the next days or weeks, but how long
it'll take will probably depend on how much effort I should spend on
cleaning up some of the improperly implemented things and try to
reimplement them properly, or just submit all the hacks I used as they
are without fixing them up, and also how much attention I should give to
conflicts with Marcus's stuff. I wonder what you guys feel about that?
Log:
Ove Kaaven <ovek@transgaming.com>
Fixed a race condition on RPC worker thread creation, and a typo.
Index: dlls/rpcrt4/rpc_server.c
===================================================================
RCS file: /cvsroot/rewind/rewind/dlls/rpcrt4/rpc_server.c,v
retrieving revision 1.12
diff -u -r1.12 rpc_server.c
--- dlls/rpcrt4/rpc_server.c 18 Apr 2003 02:04:17 -0000 1.12
+++ dlls/rpcrt4/rpc_server.c 18 Apr 2003 02:16:51 -0000
@@ -285,11 +286,18 @@
static void RPCRT4_new_client(RpcConnection* conn)
{
- conn->thread = CreateThread(NULL, 0, RPCRT4_io_thread, conn, 0, NULL);
- if (!conn->thread) {
+ HANDLE thread = CreateThread(NULL, 0, RPCRT4_io_thread, conn, 0, NULL);
+ if (!thread) {
DWORD err = GetLastError();
ERR("failed to create thread, error=%08lx\n", err);
+ RPCRT4_DestroyConnection(conn);
}
+ /* we could set conn->thread, but then we'd have to make the io_thread wait
+ * for that, otherwise the thread might finish, destroy the connection, and
+ * free the memory we'd write to before we did, causing crashes and stuff -
+ * so let's implement that later, when we really need conn->thread */
+
+ CloseHandle( thread );
}
static DWORD CALLBACK RPCRT4_server_thread(LPVOID the_arg)
@@ -315,7 +323,7 @@
}
cps = cps->Next;
}
- /* make array of connings */
+ /* make array of connections */
objs = HeapReAlloc(GetProcessHeap(), 0, objs, count*sizeof(HANDLE));
objs[0] = m_event;
count = 1;
